According to reports from multiple US media outlets including The New York Times, The Wall Street Journal, Axios, and Bloomberg, the US government indicated during a closed-door meeting with leading artificial intelligence companies on Tuesday that the frontier model cybersecurity review mechanism announced in June currently only applies to closed-source models. Open-weight models will not be affected, at least for the time being.

Insiders revealed that in the meeting, the US government defined models requiring voluntary review as closed-source models with the most advanced capabilities and national security risks. No part of this mechanism should be interpreted as restricting already released open-source models.

If an AI company chooses to submit a model to the US government for review before release, the model will enter a pre-release review period of up to 30 days. During this time, the model will undergo joint review by multiple US government agencies, and company employees will not be able to access it. The US government is not expected to publicly disclose the complete model review mechanism, but these details are not confidential, and companies that choose to participate can receive relevant detailed information.